18.14. Highly Available Storage (HAST)

Contributed by Daniel Gerzo.
With inputs from Freddie Cash, Pawel Jakub Dawidek, Michael W. Lucas and Viktor Petersson.

High availability is one of the main requirements in serious business applications and highly-available storage is a key component in such environments. In FreeBSD, the Highly Available STorage (HAST) framework allows transparent storage of the same data across several physically separated machines connected by a TCP/IP network. HAST can be understood as a network-based RAID1 (mirror), and is similar to the DRBD® storage system used in the GNU/Linux® platform. In combination with other high-availability features of FreeBSD like CARP, HAST makes it possible to build a highly-available storage cluster that is resistant to hardware failures.

The following are the main features of HAST:

After reading this section, you will know:

Before reading this section, you should:

The HAST project was sponsored by The FreeBSD Foundation with support from http://www.omc.net/ and http://www.transip.nl/.

18.14.1. HAST Operation

HAST provides synchronous block-level replication between two physical machines: the primary, also known as the master node, and the secondary, or slave node. These two machines together are referred to as a cluster.

Since HAST works in a primary-secondary configuration, it allows only one of the cluster nodes to be active at any given time. The primary node, also called active, is the one which will handle all the I/O requests to HAST-managed devices. The secondary node is automatically synchronized from the primary node.

The physical components of the HAST system are the local disk on primary node, and the disk on the remote, secondary node.

HAST operates synchronously on a block level, making it transparent to file systems and applications. HAST provides regular GEOM providers in /dev/hast/ for use by other tools or applications. There is no difference between using HAST-provided devices and raw disks or partitions.

Each write, delete, or flush operation is sent to both the local disk and to the remote disk over TCP/IP. Each read operation is served from the local disk, unless the local disk is not up-to-date or an I/O error occurs. In such cases, the read operation is sent to the secondary node.

HAST tries to provide fast failure recovery. For this reason, it is important to reduce synchronization time after a node's outage. To provide fast synchronization, HAST manages an on-disk bitmap of dirty extents and only synchronizes those during a regular synchronization, with an exception of the initial sync.

There are many ways to handle synchronization. HAST implements several replication modes to handle different synchronization methods:

  • memsync: This mode reports a write operation as completed when the local write operation is finished and when the remote node acknowledges data arrival, but before actually storing the data. The data on the remote node will be stored directly after sending the acknowledgement. This mode is intended to reduce latency, but still provides good reliability.

  • fullsync: This mode reports a write operation as completed when both the local write and the remote write complete. This is the safest and the slowest replication mode. This mode is the default.

  • async: This mode reports a write operation as completed when the local write completes. This is the fastest and the most dangerous replication mode. It should only be used when replicating to a distant node where latency is too high for other modes.

18.14.2. HAST Configuration

The HAST framework consists of several components:

  • The hastd(8) daemon which provides data synchronization. When this daemon is started, it will automatically load geom_gate.ko.

  • The userland management utility, hastctl(8).

  • The hast.conf(5) configuration file. This file must exist before starting hastd.

Users who prefer to statically build GEOM_GATE support into the kernel should add this line to the custom kernel configuration file, then rebuild the kernel using the instructions in Chapter 9, Configuring the FreeBSD Kernel:

options	GEOM_GATE

The following example describes how to configure two nodes in master-slave/primary-secondary operation using HAST to replicate the data between the two. The nodes will be called hasta, with an IP address of 172.16.0.1, and hastb, with an IP address of 172.16.0.2. Both nodes will have a dedicated hard drive /dev/ad6 of the same size for HAST operation. The HAST pool, sometimes referred to as a resource or the GEOM provider in /dev/hast/, will be called test.

Configuration of HAST is done using /etc/hast.conf. This file should be identical on both nodes. The simplest configuration is:

resource test {
	on hasta {
		local /dev/ad6
		remote 172.16.0.2
	}
	on hastb {
		local /dev/ad6
		remote 172.16.0.1
	}
}

For more advanced configuration, refer to hast.conf(5).

Tip:

It is also possible to use host names in the remote statements if the hosts are resolvable and defined either in /etc/hosts or in the local DNS.

Once the configuration exists on both nodes, the HAST pool can be created. Run these commands on both nodes to place the initial metadata onto the local disk and to start hastd(8):

# hastctl create test
# service hastd onestart

Note:

It is not possible to use GEOM providers with an existing file system or to convert an existing storage to a HAST-managed pool. This procedure needs to store some metadata on the provider and there will not be enough required space available on an existing provider.

A HAST node's primary or secondary role is selected by an administrator, or software like Heartbeat, using hastctl(8). On the primary node, hasta, issue this command:

# hastctl role primary test

Run this command on the secondary node, hastb:

# hastctl role secondary test

Verify the result by running hastctl on each node:

# hastctl status test

Check the status line in the output. If it says degraded, something is wrong with the configuration file. It should say complete on each node, meaning that the synchronization between the nodes has started. The synchronization completes when hastctl status reports 0 bytes of dirty extents.

The next step is to create a file system on the GEOM provider and mount it. This must be done on the primary node. Creating the file system can take a few minutes, depending on the size of the hard drive. This example creates a UFS file system on /dev/hast/test:

# newfs -U /dev/hast/test
# mkdir /hast/test
# mount /dev/hast/test /hast/test

Once the HAST framework is configured properly, the final step is to make sure that HAST is started automatically during system boot. Add this line to /etc/rc.conf:

hastd_enable="YES"

18.14.2.1. Failover Configuration

The goal of this example is to build a robust storage system which is resistant to the failure of any given node. If the primary node fails, the secondary node is there to take over seamlessly, check and mount the file system, and continue to work without missing a single bit of data.

To accomplish this task, the Common Address Redundancy Protocol (CARP) is used to provide for automatic failover at the IP layer. CARP allows multiple hosts on the same network segment to share an IP address. Set up CARP on both nodes of the cluster according to the documentation available in Section 30.10, “Common Address Redundancy Protocol (CARP)”. In this example, each node will have its own management IP address and a shared IP address of 172.16.0.254. The primary HAST node of the cluster must be the master CARP node.

The HAST pool created in the previous section is now ready to be exported to the other hosts on the network. This can be accomplished by exporting it through NFS or Samba, using the shared IP address 172.16.0.254. The only problem which remains unresolved is an automatic failover should the primary node fail.

In the event of CARP interfaces going up or down, the FreeBSD operating system generates a devd(8) event, making it possible to watch for state changes on the CARP interfaces. A state change on the CARP interface is an indication that one of the nodes failed or came back online. These state change events make it possible to run a script which will automatically handle the HAST failover.

To catch state changes on the CARP interfaces, add this configuration to /etc/devd.conf on each node:

notify 30 {
	match "system" "IFNET";
	match "subsystem" "carp0";
	match "type" "LINK_UP";
	action "/usr/local/sbin/carp-hast-switch master";
};

notify 30 {
	match "system" "IFNET";
	match "subsystem" "carp0";
	match "type" "LINK_DOWN";
	action "/usr/local/sbin/carp-hast-switch slave";
};

Note:

If the systems are running FreeBSD 10 or higher, replace carp0 with the name of the CARP-configured interface.

Restart devd(8) on both nodes to put the new configuration into effect:

# service devd restart

When the specified interface state changes by going up or down , the system generates a notification, allowing the devd(8) subsystem to run the specified automatic failover script, /usr/local/sbin/carp-hast-switch. For further clarification about this configuration, refer to devd.conf(5).

Here is an example of an automated failover script:

#!/bin/sh

# Original script by Freddie Cash <fjwcash@gmail.com>
# Modified by Michael W. Lucas <mwlucas@BlackHelicopters.org>
# and Viktor Petersson <vpetersson@wireload.net>

# The names of the HAST resources, as listed in /etc/hast.conf
resources="test"

# delay in mounting HAST resource after becoming master
# make your best guess
delay=3

# logging
log="local0.debug"
name="carp-hast"

# end of user configurable stuff

case "$1" in
	master)
		logger -p $log -t $name "Switching to primary provider for ${resources}."
		sleep ${delay}

		# Wait for any "hastd secondary" processes to stop
		for disk in ${resources}; do
			while $( pgrep -lf "hastd: ${disk} \(secondary\)" > /dev/null 2>&1 ); do
				sleep 1
			done

			# Switch role for each disk
			hastctl role primary ${disk}
			if [ $? -ne 0 ]; then
				logger -p $log -t $name "Unable to change role to primary for resource ${disk}."
				exit 1
			fi
		done

		# Wait for the /dev/hast/* devices to appear
		for disk in ${resources}; do
			for I in $( jot 60 ); do
				[ -c "/dev/hast/${disk}" ] && break
				sleep 0.5
			done

			if [ ! -c "/dev/hast/${disk}" ]; then
				logger -p $log -t $name "GEOM provider /dev/hast/${disk} did not appear."
				exit 1
			fi
		done

		logger -p $log -t $name "Role for HAST resources ${resources} switched to primary."


		logger -p $log -t $name "Mounting disks."
		for disk in ${resources}; do
			mkdir -p /hast/${disk}
			fsck -p -y -t ufs /dev/hast/${disk}
			mount /dev/hast/${disk} /hast/${disk}
		done

	;;

	slave)
		logger -p $log -t $name "Switching to secondary provider for ${resources}."

		# Switch roles for the HAST resources
		for disk in ${resources}; do
			if ! mount | grep -q "^/dev/hast/${disk} on "
			then
			else
				umount -f /hast/${disk}
			fi
			sleep $delay
			hastctl role secondary ${disk} 2>&1
			if [ $? -ne 0 ]; then
				logger -p $log -t $name "Unable to switch role to secondary for resource ${disk}."
				exit 1
			fi
			logger -p $log -t $name "Role switched to secondary for resource ${disk}."
		done
	;;
esac

In a nutshell, the script takes these actions when a node becomes master:

  • Promotes the HAST pool to primary on the other node.

  • Checks the file system under the HAST pool.

  • Mounts the pool.

When a node becomes secondary:

  • Unmounts the HAST pool.

  • Degrades the HAST pool to secondary.

Caution:

This is just an example script which serves as a proof of concept. It does not handle all the possible scenarios and can be extended or altered in any way, for example, to start or stop required services.

Tip:

For this example, a standard UFS file system was used. To reduce the time needed for recovery, a journal-enabled UFS or ZFS file system can be used instead.

More detailed information with additional examples can be found at http://wiki.FreeBSD.org/HAST.

18.14.3. Troubleshooting

HAST should generally work without issues. However, as with any other software product, there may be times when it does not work as supposed. The sources of the problems may be different, but the rule of thumb is to ensure that the time is synchronized between the nodes of the cluster.

When troubleshooting HAST, the debugging level of hastd(8) should be increased by starting hastd with -d. This argument may be specified multiple times to further increase the debugging level. Consider also using -F, which starts hastd in the foreground.

18.14.3.1. Recovering from the Split-brain Condition

Split-brain occurs when the nodes of the cluster are unable to communicate with each other, and both are configured as primary. This is a dangerous condition because it allows both nodes to make incompatible changes to the data. This problem must be corrected manually by the system administrator.

The administrator must decide which node has more important changes or merge them manually. Then, let HAST perform full synchronization of the node which has the broken data. To do this, issue these commands on the node which needs to be resynchronized:

# hastctl role init test
# hastctl create test
# hastctl role secondary test

All FreeBSD documents are available for download at http://ftp.FreeBSD.org/pub/FreeBSD/doc/

Questions that are not answered by the documentation may be sent to <freebsd-questions@FreeBSD.org>.
Send questions about this document to <freebsd-doc@FreeBSD.org>.